Google-Apps
Hauptmenü

Post a Comment On: Ken Shirriff's blog

"Iconic consoles of the IBM System/360 mainframes, 55 years old"

36 Comments -

1 – 36 of 36
Anonymous Frank Littler said...

I joined IBM in late 1976 and was surprised how physically large the machines were.they consumed lots of electricity and gave out a lot of heat hence the need for water cooling.But considering their complexity they were reliable and easily fixable quickly when something went wrong.
The company almost went bust in later years by trying to sell their own old technology and following the wrong one and so had some catching up to do. Head in the sand springs to mind. Their management systems could not keep up with the success of the PC.

April 7, 2019 at 12:23 PM

Blogger Mike007 said...

I was surprised to find this 'old computer' running the UK's air traffic control system when I joined the CAA in 1989 - it used an IBM 9020D which is a triplex (3) model 65 S/360 with two model 50s as I/O 'elements' for radar data processing. It was replaced the same year by dual IBM 4381s; gone were the card reader and ounch, though we continued to use mag tape.

April 7, 2019 at 12:44 PM

Blogger Michele said...

Thanks for the trip down memory lane. My university threw a big party, complete with cake and party hats when they expanded the 360/65's memory from 256K to a whopping 512K. The 360/65 was also capable of time-sharing, via the TSO option.

We had dozens of Selectric terminals and CRT's available across campus. Heady stuff back in the early 70's. Of course time-sharing access entailed an extra charge.

April 7, 2019 at 5:04 PM

Anonymous George Haeh said...

As a system programmer, I was hands on with several 360/70 consoles and control panels from 30s that I managed to stuff OS/360 HASP into up to 165s?. PCP, MFT, MVS,VM and guess what? No Viruses!

April 7, 2019 at 7:28 PM

Blogger Scott Sherrell said...

Another wonderful blog post, Ken. Thank you!

How might a planning project team have been able to determine the costs associated with the environmental maintenance and electrical costs that a given system would require, on top of the monthly rental amount? Because I am assuming that a business would have some idea of the workload that would be performed, would there have been a way to compare and contrast the total cost of different models given that same load?

I’m completely fascinated with trying to get my mind around how these systems were implemented and how companies could calculate their returns on the investment.

April 7, 2019 at 9:13 PM

Anonymous Yair E said...

Thanks for this beautiful post.
One of the things I remember from my MF times is that the documentation specified how many master-terminal operators, system programmers, application programmers etc you needed to run a certain work load...

April 8, 2019 at 3:44 AM

Anonymous Mike Ross said...

Ken, while this is a fantastic resource, I feel it would add to the interest to embed videos of these consoles in operation, where possible. I know LCM+L have an operational 2030, but I was also thinking of existing YouTube videos of LCW's 2030 console, driven by a complete emulation of the 2030 hardware:

https://www.youtube.com/watch?v=_8YXFhowc3A

and Camiel Vanderhoeven's working 2065 console emulation, which is unfinished but starts to IPL OS/360:

https://www.youtube.com/watch?v=Fm-ZMACqrK8

Cheers

Mike


April 8, 2019 at 5:23 AM

Blogger Frank Baldino said...

Great job Ken. I was trained on the 360 Model 25 in 1976 as an IBM Customer Engineer. It was used as the front end to the 3890 check sorter. The early versions used core memory and I believe that may be the last storage technology where you can visually see 1 bit. I still have a 2K module consisting of several flat planes. There are 9 'donuts' for each the byte, the last bit serving as an error detection bit. These machines were really a lot of fun to diagnose and fix.

April 8, 2019 at 6:55 AM

Blogger Karol Kasanicky said...

Thank you, really very interesting blog. In these times, we were isolated behind "iron curtain", first IBM machine I was working on was 370/145.
Thank you once more.

April 8, 2019 at 10:47 AM

Blogger WmHBlair said...

| 15. A CRT display was used for the console on
| the models 85, 91 and 195. (I suspect this was
| because CRT display technology had advanced by
| the time these later systems were built.)

It had nothing to do with CRT display technology
(advanced OR cheaper). In the case of the 360/91
(and 360/95) and the 360/195 (and 370/195), the
reason for a CRT operator console was speed; the
"printer" (and Selectric typewriter ball-based)
consoles were simply too slow for the fast CPUs.
The 91 (and 95) and 195 models still had huge,
control and indicator panels, but they were not
integrated with their CRT-based operator display
consoles. The 360/85 used an integrated system
console that incorporated both the CRT console
display and the indicator and control panel, but
the indicator panel section was small [like the
CRT console] and the control panel section was
very small (smaller than the one for the 360/30)
and integrated with the chassis that housed the
CRT display.

In other words, starting with the 360/85, they
stopped building huge "blinking lights" panels
that were so popular with executives, visitors,
and media photographers. The CRTs were needed
for message display speed, regardless of cost
(or technology), and computer rooms needed the
space occupied by the 6-, 7-, and 8-foot wide
blinking lights and control panels for actually
useful equipment.

The size of the control and display panels (as
distinguished from the actual "operator console"
on which messages were typed or displayed) was
merely a reflection of the internal complexity
of the CPU. The knobs, switches and blinking
lights were of essentially no use to the actual
computer operator (other than impressing visitors
and enabling one to observe patterns indicating
that the system was operating normally -- or not).
They were mainly needed by the IBM FEs performing
maintenance and running diagnostics.

The indicator panels for the 91 (and 95) and the
195 were huge simply because there was a lot that
needed to be displayed. The Model 360/85 was more
complex; an indicator and control panel for it
that used "discrete" lights and switches would
have been unacceptably huge, so they used the
"Microfiche Indicator Viewer" (part of the "System
Control Panel") instead. The actual 360/85 operator
console (the "Main Control Panel") housed the CRT
(which displayed messages issued by the operating
system, etc.) and the knobs, buttons, and switches
actually used by the operator (such as the blue
LOAD/IPL button and the Load Unit Address dials).

This 360/85 "System Control Panel" was also used
for the System/370 Models 165 and 168 (where IBM
called it the "IBM 3066 System Console"). As far
as I was able to tell in 1971, the 3066 was only
cosmetically different from the 360/85 "console."

April 8, 2019 at 11:45 AM

Blogger WmHBlair said...

| The models 91 and 195 used a display based on
| the IBM 2250 Graphics Display Unit. This was a
| vector display, drawing characters from line
| segments instead of pixels.

The integrated operator console on the 360/85 was
also a vector display, and was also based on the
IBM 2250, except I don't remember it having a light
pen. The System/370 Model 165 was essentially the
same machine as the 360/85, except that it used
"old" (used, returned) 360/50 core memory modules,
which caused many legal and contractual problems,
because the 370/165 was being sold as (all) "new"
and many organizations and state and federal tax
laws treated (genuine) new equipment differently.
I was a hands-on user of both the IBM 2250 display
and the 370/165 operator console.

The operator console on the 360/195 (and 370/195)
was called the "IBM 3060 Model 1 System Console."
It was also based on the IBM 2250 Graphics Display
Unit, but was, IMHO, much better (more lines could
be displayed). The IBM 3066 console used with the
various 370/165 and 370/168 models, along with the
earlier, nearly-identical twin, the 360/85 System
Control Panel, was smaller (about half as tall and
showing about half as many lines of text) than the
360/91 (and 360/95) and 360/195 (and 370/195) CRT
display consoles; the vector graphic characters
were clunkier, larger, and seemed to be drawn in
a sloppy manner (line segment ends did not always
match up nicely), and very much less clear than
the 3060 (and 2250) CRT characters. One would have
thought they would be as well rendered (or even
better, for supposedly more modern technology),
but the 3066 definitely had a low-budget look.
That said, the 3066 was FAST (which goes a long
way towards explaining the difference I suspect).

April 8, 2019 at 11:45 AM

Blogger David Cortesi said...

As a hardware CE in the late 60s I was called in from fixing unit-record gear to help with the installation of the first model 85 in the San Francisco Data Center, which was an all-hands-on-deck effort for the local branch office. I recall the steel frames being installed then lots of copper piping for the coolant system being plumbed into them.

I also recall how the weakest link in all of these systems was the console printer. It was a beefed-up Selectric typewriter but even with improved bearings, the Selectric mechanism was never meant to pound away printing OS messages constantly 24/7. When the console printer failed, DOS or MVS stopped and nothing could be run until it was fixed. So the CE who had Selectric training was liable to late-night callouts at any time.

April 8, 2019 at 12:14 PM

Blogger Zom-B said...

The images in the article itself are blurry and upscaled (on my HiDPI screen) so I thought the originals were just unfortunately crappy. Until you talked about a 'flow diagram [of the Model 50]', which I couldn't spot, until... snap, the images are clickable! And are actually of a higher resolution.

As almost all webpages I know don't display upscaled thumbnails or images anymore (i.e. they're HiDPI compatible), I wasn't actively paying attention to it. I don't know if it's the fault of Blogger or how you upload the images (as I don't know other Blogger sites)

April 9, 2019 at 10:55 PM

Blogger AreTwo said...

Love your work. :-) I have used a lot of these systems since 1966 when I entered The Computer scene. We ordered a "40" and time on a "30" until it was delivered.
I would love similar historical record for the IBM operating systems.
In 1966, we ran BPS for Fortran as the "new" Fortran 4 would not run on BOS, and Basic Operating System (BOS), both Tape and Disk resident for accounting and managerial applications. Then BOS became TOS or DOS, then the HUGE upgrade - DOS 10! etc etc..

April 10, 2019 at 4:50 PM

Blogger Carl Claunch said...

The planning guides for the machines gave power consumption figures - so many KVA for box 1234 and feature abcd. They also listed heat output which allowed the installtion to figure out the power requirements for both powering the machine and cooling it.

Given those levels. the customer could get pricing on so many tons of cooling and so many A of 3 phase power, for the ACs and the electrical conditioning, switching and wiring costs.

A customer knew quite accurately what it would cost to build up the data center and then to operate the machine,

The final elements were the labor costs of operators and systems programmers, plus the maintenance monthly charges to IBM.

April 10, 2019 at 6:55 PM

Blogger Kaleberg said...

Thanks for a great rundown of the 360 series. It was a classic, not just of technology, but of marketing design. Alfred Sloan built GM to produce a car for customers in narrow income bands, so GM had many car brands and models within those brands. IBM segmented its market similarly, and it really shows with the 360.

I saw the 360/95 at Columbia/Lamont-Doherty's and NASA's Goddard Center back around 1970. It had a huge 4MB memory module. It was the size of an industrial AC unit with water pipes for cooling running in and out of it. I was told it was all core memory. Was it replaced with a thin film memory after my visit? Was I listening to an unreliable narrator? He went on to head computing at the NSA.

Between that memory unit and the CPU there was a much smaller unit placed in a space clearly allocated for a module commensurate with the core memory unit. I was told it was a memory cache unit using "monolithic" memory technology. I was in high school and technologies were in flux, so I never got more detail than that. Perhaps that was the thin film memory unit? I vaguely remember it as having a 16KB capacity, but this is a rather vague memory.

April 14, 2019 at 10:44 AM

Blogger WmHBlair said...

Kaleberg stated/asked:

| I saw the 360/95 at Columbia/Lamont-Doherty's and NASA's
| Goddard Center back around 1970. It had a huge 4MB memory
| module. It was the size of an industrial AC unit with water
| pipes for cooling running in and out of it.
| I was told it was all core memory.
| Was it replaced with a thin film memory after my visit?
| Between that memory unit and the CPU there was a much smaller
| unit placed in a space clearly allocated for a module
| commensurate with the core memory unit. I was told it was a
| memory cache unit using "monolithic" memory technology.
| Perhaps that was the thin film memory unit?

Kaleberg, Your memory is excellent. To answer your questions:
The 4MB core memory unit was NOT replaced after your visit.
The 360/95 had BOTH core and thin-film memory units. Each
thin-film memory unit contained 64KB; the 360/95 had 16 thin-
film memory units totaling 1024KB (1MB) of very fast memory.
In addition, the 360/95 had 4MB of (ordinary) core memory.
Hence, a total of 5MB of "main" memory: 1MB really fast and
4MB relatively slow (but nonetheless very fast compared to
all but the 360/65 and 360/75 [and possibly the 360/85]).

April 23, 2019 at 12:27 PM

Comment deleted

This comment has been removed by a blog administrator.

April 24, 2019 at 2:54 PM

Anonymous Anonymous said...

Many Thanks for this great page!
I started my IBM career in 1979 at the Munich branch office. We had a /360 Mod. 91 installed at our customer 'Institute for Plasma Physics' (IPP). Your picture 'Console of the IBM System/360 Model 91' is NOT the Munich machine.
https://www.mpcdf.mpg.de/about-mpcdf/publications/bits-n-bytes?BB-View=200&BB-Doc=211
The Mod. 20 was developed by the German IBM laboratory in Boeblingen, this lab as a collection of historic IBM products, including an Mod. 20 which is still in running condition. The German Lab is still working on new generations of IBM System z mainframes.
One Mod. 20 was rescued from defunct German factory in April 2019, see the great story
https://ibms360.co.uk/?page_id=22
best regards
Norbert Ziegeler

August 7, 2019 at 3:13 AM

Blogger PfJ said...

As an ex-IBMer (40 very happy years) who started with the 1401 and 360/30 in 1970 I’d just like to point out one factual error ... we never “booted” a 360 (or 370 later) ... we I...P...L...ed or re-IPLed the system. The IPL record was placed upon Cylinder 0 Track 0 of the IPL device which the channel program read when the IPL button was pressed.

Although it was also possible to IPL from a deck of punch cards (typically 00C), paper tape (typically 007), or magnetic tape (typically 181), if you needed to perform some maintenance on the system disk.

The expression Boot was not used until the part of the system disk from which the system was started was contained in a Boot Sector which came about in 1981 with the IBM PC [although it’s possible that the term Boot was used in IBM office products ... a part of IBM with which I had no familiarity].

November 9, 2019 at 1:54 AM

Blogger AreTwo said...

PJ, IPL stands for Initial Program Loader, and when the LOAD button was pressed the program it executed was called the "Bootstrap" program - A bootstrap was the little loop at the back of a boot to pull it on by, so a bootstrap program was a little program that got your program in and started.
Other manufacturers did not have an IPL button like IBM, and I've even had to do an IPL by actually keying the bootstrap program via front panel keys!
Usually the bootstrap program consisted of two instructions - Read 1 record from the boot device into a memory location then branch to that location. Therefore that record had to contain the "real" loader program, which would then load and execute your program.

November 9, 2019 at 4:02 PM

Blogger Jim St said...

At the University of Missouri, Rolla (now MST in Rolla), we had a 360/50. It ran a subsystem called CPS, which was a firmware extension supported timeshare system, as well as running as a 360/50 running MVT 21 and HASP.

The CPS system included an outboard memory system called LCS (don't have a number), which had 1mb of memory, and when CPS was up, was used to support the users.

CPS could run with the MVT system active.

Our system had 512K of memory in the main box.

Two unique things about the 360/50 at UMR. It was supposedly the only non power of 2 system ever built up by IBM which could run as such. 1.5mb was not a standard for 360 memory sizes.

And supposedly, it was also the only non leased 50 at the time. IBM sold the system during the development of the 360/50 and shipped a 360/40 to the campus in 1969 or 1970 which was used for around a year till the /50 could be delivered to replace it.

Jim Stephens

January 17, 2020 at 3:58 PM

Blogger WmHBlair said...

| Two unique things about the 360/50 at UMR. It was supposedly the only
| non power of 2 system ever built up by IBM which could run as such.
| 1.5mb was not a standard for 360 memory sizes.

Nope. There were many of those. I had one that was 256KB (fast core) + 1MB (LCS). There were others I knew of (and just in NC), including 1.5MB (512KB regular + 1MB LCS). OS/360 NIP recognized these "odd" main core storage sizes and could figure out how much was "fast" and how much was "slow" (i.e., LCS). Unmodified, all the LCS core went into "Hierarchy 1" storage.

| IBM sold the system during the development of the 360/50

That would have been around 1965 (not 1969).

| and shipped a 360/40 to the campus in 1969 or 1970

By then, there were (I suspect) hundreds of 360/50 installations. It was a very popular machine (much more flexible than a /40).

| which was used for around a year
| till the /50 could be delivered to replace it.

It is much more likely that your 1969 or 1970 date is incorrect.

There were, in fact, "many" 360/40 machines shipped to customers who wanted larger ones (including 360/75s) to use temporarily until their box could be delivered. But all of those production problems were over by 1967 and everybody that wanted a /50 or /65 or /75 had them or could get them (with the usual sort of order lead time that was expected then).

I also used CPS, and loved it for some interesting research applications. It was actually much better than TSO in OS/360 Release 20.1 and 21.x. IMHO, TSO was essentially useless except for limited applications [and this does not include text or program editing) until the OS/VS2 1.6 and SPF era arrived.



February 1, 2020 at 11:37 AM

Blogger Len said...

There were 2 versions of the 360/50 front panel. See http://www.quadibloc.com/comp/pan04.htm for both. Ones with smaller main storage had a very cute diagram of the CPU dataflow on the top right.

Thanks for the great site-pictures!

Regards, Len

March 23, 2020 at 12:52 AM

Anonymous Anonymous said...

I still work with the main UK Air Traffic control system - originally written for 6 x 360-65 and now very happy on a a couple of Z-Series processors. Such elegant programming and resilience built in the architecture.

April 16, 2020 at 1:29 PM

Blogger WmHBlair said...

The link http://www.quadibloc.com/comp/pan04.htm is broken or compromised (it leads to an insecure location that cause several malware detection tools to call a halt to proceeding further).

April 17, 2020 at 3:11 PM

Blogger Unknown said...

I joined LJ Hooker as a Programmer/operator on 3rd September 1965 in Sydney Australia and we got one of the first 6 that were shipped into Australia. Prior to that we were programming, in Assembler and testing on IBMs machine down at the Pagoda near the Sydney Harbour Bridge. Our science teacher from my high school the previous year was teaching us programming and begged me not to tell anyone because he was only a few pages in front of us in the reference manual. The 1311 disk drives only held 11 million characters and we had to fit programmes into 32k. We had no console typewriter so we had to dial instructions in. The following year I joined an aftermarket company and when a customer came to the front desk to enquire after a part, we had to pull the system interrupt to stop the computer, load the enquiry in on a card and check the print out from the printer.
We found the disk drives were great for sleeping behind when we pulled the 24 hour shifts to implement systems. They threw out a lot of heat in that air conditioned room.
We definitely felt like pioneers.

July 7, 2020 at 1:53 AM

Anonymous Data Integration Expert said...

We have come a long way. From hardware driven enterprises to data-driven enterprises. Companies today are much mroe flexible, have a ton of data available and software centric. Today, hardware is not a problem but speed and processing power is!

September 16, 2020 at 9:30 AM

Anonymous Anonymous said...

I tripped across this blog when searching for an old IBM part. Glad to see the information captured and catalogued in this very complete form. Many thanks.

I joined IBM in 1967 (it was referred to as the "Gold Rush" period) as a Field Engineer. I specialized in the Model 50, then later in the S370 155, 155DAT, and 158. Lots of fun and lots of great memories.

I worked far deeper in these systems than most IBMers. If you want to read a bit more "inside" information, pick up a copy of my life story (or at least the first 70 years) titled "From Shasta to Shanghai". It is available on Amazon.

Tim Griswold
tgriswold@live.com

November 8, 2020 at 11:34 AM

Blogger Jim Snellen said...

A very good read Ken! It was a real treat to revisit my days as an operator of the 360/20. It was my very first mainframe experience back in 1972. I then went through several models of the System 370 line and then settled on the 4331, 4341, and finally a 4381 in the late1990’s. Thanks for the memories.

December 14, 2020 at 6:29 AM

Blogger Ken Freeman said...

My first exposure to the 360 line was as an operator of a 360/30 on the night shift at a bank. Later, wrote my first COBOL programs on a 360/40. A great, nicely done article!

December 27, 2020 at 9:06 AM

Anonymous HeckSpawn said...

Learned my first bit of computing on a 360/40 at the Computing Learning Center of Los Angles with a dos/vse system. Said goodnight to what was probably the last dos/vse system on 21/dec/09 at EDS with one of the last OS's still running it for Phillips.

January 6, 2021 at 3:14 AM

Anonymous Kevin G. Rhoads said...

A borrowable version of IBM's 360 and Early 370 Computers is available at
https://archive.org/details/ibms360early370s0000pugh

Also the history of the predecessors

IBM's Early Computers
https://archive.org/details/ibmsearlycompute00bash

April 15, 2021 at 3:25 PM

Blogger Skylab said...

Excellent post Ken. We loved our "Big Iron". I started my operator career at a Texas Junior-College on a mod40, parlayed that into a gig with the State of Texas herding "jobs" through a mod50 + mod40 pair on grave-yard shift, and was tickled pink to land a position with Lockheed-Electronics operating in the five-plex of mod75RT behemoths sporting LCS (Real-Time mods included the clock [of course] and circuitry for four additional interrupts plus the necessary new machine instructions to handle those interrupts, with extensions to the PCP Kernal under OS360RT the beast ran) that NASA used for the Apollo program. We "flew" 16 & 17 and I hung around until the Apollo-Skylab program shutdown. By the early '90s, I'd completed the programmer-analyst-manager progression out to the end of my hands-on main-frame career. The last 25 years have been internet-centric but you might be surprised how often my main-frame chops have come in handy...

May 17, 2021 at 12:47 PM

Blogger Rob said...

Hey all-


I have a control panel of a model 40 in my possession, looking to find a missing button and “red emergency off”

Anyone with info please reach out!

January 16, 2022 at 11:45 AM

Blogger John W said...

You missed two models—the 64 and the 66. They were variations on the 60 and 62, with DAT (Dynamic Address Translation—virtual memory) added. You missed them because it was only one month later that the 60 and 64 were dropped altogether, and the 62, 66, and 70 were replaced with the 65, 67, and 75.

There was also the 69, but that was a joke, with added op-codes such as Halt and Catch Fire, Punch Disk, and Punch Operator.

(There was also a joke about “OS/360: Multiprogramming with a Variable Number of CPUs”, but, dear God, we actually have that now!)

February 20, 2022 at 10:09 AM

You can use some HTML tags, such as <b>, <i>, <a>

Comment moderation has been enabled. All comments must be approved by the blog author.

You will be asked to sign in after submitting your comment.
Please prove you're not a robot